Starting from the H.264 highly compressed digital video codec standard, intra-frame direction prediction coding has become the mainstream technology of video intra-frame block coding. In such technique, the prediction of the current coding block is linearly predicted from its left adjacent column and upper adjacent row according to one direction. As shown in FIG. 1, the current coding block is an 8×8 block, and the pixels represented by the triangle are the left and upper adjacent reference pixels of the current block. Intra-frame prediction first specifies a direction, and then each row or column in the current block is obtained from its left adjacent column and upper adjacent row according to this direction, the sixth row as shown in the figure is mapped to the pixel position of the upper adjacent row according to the direction shown in the figure, as shown by the dots in the figure. Because these dots are not the integer pixel, they need to be calculated based on their subpixel coordinates by interpolation algorithm according to the known integer pixel, that is, the triangle pixels shown in the figure.
One problem presents in the existing intra-frame prediction is that, since the reference prediction pixel is located at the left and upper adjacent positions of the block, and the prediction correlation is lower as the distance is farther, so for the right and lower pixel of the current block whose distance is farther, it is clear that the prediction accuracy is not high, thus reducing the coding efficiency.
How to provide a video coding intra-frame prediction method and video coding intra-frame prediction device of high prediction accuracy have become an urgent problem to be solved.